In this paper, we take the mean and standard deviation of the RCS frequency domain fluctuations was taked asthe recognition feature, the adaptive radial basis network based on sample density was used to do research on ship recogni.tion and classification. The improved adaptive PSO method was used to estimate the optimal neighborhood radius of sam.ple density, which realized the adaptive selection of radial basis network center. The improved adaptive PSO method im.proved the search speed and precision of the neighborhood radius of the optimal sample density. It took BWP as the fitnessevaluation function, which could reflect sample clustering features, and adopted Gauss adaptive inertia weight adjustmentstrategy. The experimental results showed that the adaptive radial basis network could adaptively obtain the RBF centerand its position distribution corresponding to the optimal identification rate of radial basis network. Therefore, it reducedthe dependence on the experience of modeling person, and enhanced the ability of the anti-ship missile to recognize thetype of ship.